Search Results (14,186)

Radar-based human activity recognition (HAR) has emerged as a promising alternative to vision-based systems, as it can operate under poor lighting, occlusions, and privacy-sensitive scenarios. However, existing radar HAR methods often suffer from limited feature extraction capability due to noise, signal attenuation, and the challenge of capturing both global motion patterns and local micro-Doppler dynamics simultaneously. To address these issues, this paper proposes RepHARNet, a novel HAR network built on structural reparameterization for radar micro-Doppler signals. Specifically, RepHARNet decomposes the weight matrices into global and local components through a reparameterization strategy, enabling the network to simultaneously capture coarse-grained inter-channel dependencies via a Global Perceptron and fine-grained intra-channel spatial correlations via a Channel Perceptron. Furthermore, a parameter-efficient share-set mechanism is integrated into the Channel Perceptron to substantially reduce the computational overhead while maintaining the representational capacity. Extensive experiments on the public IMG848 dataset demonstrate that all four RepHARNet variants achieve top-1 accuracies above 93.00%, among which RepHARNet-large achieves the highest at 94.86%, significantly outperforming existing mainstream methods. Additional evaluations on the Ci4R dataset further verify the robustness and effectiveness of RepHARNet, where all variants achieve competitive accuracies above 90.00%. The results verify the effectiveness and superiority of RepHARNet in radar-based HAR. Full article

Clarifying the mutual feedback relationship and coordinated evolution characteristics between low-carbon transportation (LCT) and high-quality economic development (HQED) is of great significance for the green transformation of resource-based and ecologically fragile urban agglomerations. Taking 18 cities in the Yellow River Jiziwan Metropolitan Area as the research objects, this paper constructs an evaluation indicator system for LCT and HQED based on panel data from 2013 to 2022, and comprehensively applies the ISM-MICMAC model, a modified coupling coordination degree model, a gravity model, an obstacle degree model, and a combined GM-ARIMA forecasting model to analyze the interaction relationships, spatiotemporal evolution, spatial correlations, and scenario differences between the two systems. The results indicate that: (1) A hierarchical mutual feedback relationship exists between LCT and HQED, in which the relevant factors exhibit a hierarchical association within the system structure, extending from basic input, transportation supply, and economic operation to green and low-carbon outcomes. (2) During the study period, the comprehensive development levels of the two systems generally improved, with the mean coupling coordination degree rising from 0.4374 in 2013 to 0.4702 in 2022, remaining overall at a borderline coordination stage, while inter-city divergence was relatively pronounced. (3) The spatial connection network gradually exhibited multi-node linkage characteristics, yet strong connections remained concentrated in a few core cities. (4) Scenario predictions reveal that the synergistic development scenario is most conducive to enhancing the coupling coordination level, and the differences among scenarios gradually widen after 2026. Simultaneously advancing LCT and HQED is an important pathway to enhance the regional synergy level of the Yellow River Jiziwan Metropolitan Area. Full article

In low-light, occluded, and cluttered environments, single-modality RGB detectors are prone to false positives and missed detections. While infrared (IR) imaging provides relatively stable target visibility under poor illumination, it lacks texture and color information and is susceptible to background thermal noise and imaging variations. To address these limitations, this paper proposes an RGB–IR object detection network, named MCF-YOLO, consisting of three core components. First, the Cross-Modal Hierarchical Fusion (CMHF) module performs stage-wise alignment and fusion on multi-scale features, jointly modeling RGB texture details and IR thermal responses to exploit the structural and semantic complementarity between the two modalities. Second, the Soft Attention Regularization based on Attention Prior (SAR-AP) module derives attention priors from IR features to impose soft constraints on cross-modal attention maps. This mechanism helps the network maintain attention on target-relevant regions, thereby suppressing attention drift caused by low-light noise and complex backgrounds. Third, the Small-Object-Sensitive Detection Head (SOS-Head) processes high-resolution features to strengthen the representation of small targets, improving detection capability in long-range and occluded scenarios. In evaluations on two RGB–IR benchmarks—M3FD and VEDAI—MCF-YOLO achieves improvements of 2.7% in mAP@0.5 and 1.1% in mAP@0.5:0.95 on M3FD, and 5.4% and 4.4%, respectively, on VEDAI. These results suggest that consistency-guided cross-modal fusion and high-resolution small-target modeling are beneficial for RGB–IR detection in low-visibility and cluttered scenes. Full article

In the fabric warehouse, order picking operations face high labor intensity and rising operational costs, requiring urgent optimization. This study investigates the order scheduling and task assignment problem within an elastic staffing framework, where temporary labor recruitment and real-time task allocation need to be adjusted dynamically in response to fluctuations in order volumes. Nevertheless, conventional approaches often suffer from severe computational bottlenecks under such highly dynamic conditions, and struggle to maintain optimal solutions when demand undergoes large and frequent fluctuations. To address these challenges, this study proposes a Graph Transformer Policy Network with Proximal Policy Optimization (GTP-PPO), which combines graph structure features with a global attention mechanism. First, the return picking strategy and the S-shaped picking strategy are compared and analyzed in the fabric warehouse scenario. The results reveal that the return strategy is more suitable for the studied warehouse layout. Subsequently, a mixed-integer programming (MIP) model and a GTP-PPO model are established for optimizing order dispatching and scheduling. Finally, an empirical analysis is carried out based on the peak order day of the year in the fabric warehouse. The results demonstrate that the proposed GTP-PPO model not only achieves near-global optimal solutions (gap < 4%) comparable to the MIP model, but also exhibits robust real-time decision-making capabilities under dynamically increasing order volumes and unexpected disruptions. Compared to the MIP model, the GTP-PPO approach reduces unskilled labor hours by 84.80% and decreases operational volatility by 27.60%, with only a 3.52% increase in operational costs. Full article

To address the limitations of vision-only defect detection, image–semantic misalignment, and spatial-logic conflicts in complex substation inspection scenarios, this paper proposes a camera-sensor-based multimodal defect sensing framework with cross-modal feature mapping for substation equipment monitoring. The proposed framework integrates field inspection images acquired by camera sensors, defect textual descriptions, and equipment topology knowledge and establishes a unified domain-adaptive pre-training–bidirectional cross-modal mapping–hierarchical reasoning workflow. First, a Contrastive Language–Image Pre-training (CLIP)-based domain-adaptive pre-training strategy is developed to enhance the representation of equipment categories, defect attributes, and inspection-scene semantics. Second, a bidirectional cross-modal feature mapping network is constructed to model fine-grained interactions between candidate visual regions and textual semantics, where uncertainty-aware fusion and prototype constraints are introduced to improve semantic alignment and defect discrimination. Third, a hierarchical neuro-symbolic reasoning module incorporates equipment topology and spatial rules for posterior verification, logical consistency checking, and false-positive suppression. Experiments on a substation inspection image dataset demonstrate that the proposed method achieves 90.8% mAP@0.5, 68.7% mAP@0.5:0.95, and 89.4% F1-score, outperforming mainstream and recent detection models. Full article

In this paper, a novel approach is presented for fusing RGB and LiDAR inputs for semantic segmentation. Accurate building detection is required for various scenarios such as urban planning or environmental monitoring. The two main sources for accurate building segmentation are either RGB aerial images or LiDAR point clouds covering the selected area. Each of these sources has its own well-known techniques for segmentation; however, for the combination of the input, there are not many architectures available, and extracting different features from the two different fields can result in an enhanced segmentation map. The authors of this article created a semantic segmentation model that uses both the aerial RGB image and the LiDAR point cloud as its input. The network first takes the point cloud and forwards the processed projection to a modified U-Net-based architecture, which fuses the extracted features of the 3D input with the extracted information of the 2D input on each level of the decoding. To train and test the presented model, the authors used a dataset containing more than 3000 images and their corresponding 3D point clouds of three different areas from Hungary. As is also presented in this paper, this approach provides significantly better results than the traditional RGB, Point Cloud segmentation models, and their ensembles in terms of segmentation accuracy. Full article

Deep reinforcement learning (DRL) has shown considerable potential in local path planning for autonomous robots. However, existing DRL methods still suffer from limited training efficiency, poor generalization, and weak sim-to-real transferability in complex environments. To address these issues, this paper proposes a Multi-Level Attention Dueling Double Deep Q-Network (MLA-D3QN) framework, which progressively enhances feature extraction, spatial perception, and modality fusion through three attention levels: rule-based attention for obstacle contour extraction, implicit neural multi-scale spatial attention for environment perception, and bidirectional cross-attention for multi-modal feature alignment. Simulation results show that MLA-D3QN outperforms baseline and comparison methods in terms of convergence speed and average reward. Real-world experiments are conducted on a Scout mini platform with 50 trials in simple task scenarios (sparse obstacles, short distance) and 50 trials in complex task scenarios (dense obstacles, long distance). The proposed method achieves success rates of 98% in simple tasks and 94% in complex tasks. Compared to CNN-D3QN and D3QN, MLA-D3QN improves success rates by 10 percentage points (vs. CNN-D3QN) and 38 percentage points (vs. D3QN) in simple tasks, and by 34 percentage points (vs. CNN-D3QN) and 84 percentage points (vs. D3QN) in complex tasks. Path costs are reduced by 24.0% (vs. CNN-D3QN) and 59.9% (vs. D3QN). These results validate the effectiveness of MLA-D3QN in improving generalization and sim-to-real transferability for local path planning in complex environments. Full article

Drainage pipe defect detection is essential for maintaining the normal operation of urban infrastructure. In recent years, deep learning-based object detection methods have provided an effective technical solution for drainage pipe defect recognition. Among them, YOLO-series models have demonstrated strong potential in visual detection tasks due to their end-to-end architecture and high inference efficiency. However, directly applying baseline YOLO models may still face challenges such as limited detection accuracy, relatively high model complexity, and insufficient adaptability for lightweight deployment scenarios. To address these issues, this paper proposes a lightweight drainage pipe defect detection method based on an improved YOLO11 network. Rather than treating detection enhancement and model compression as two separate procedures, the proposed method integrates feature enhancement, adaptive pruning, and distillation-based recovery into a unified lightweight detection framework. Specifically, an improved SimAM attention mechanism is introduced into the backbone and integrated with the C3k2 module to construct the C3K2_SWS module, aiming to enhance the representation capability of critical defect features. In the neck network, a focused diffusion pyramid network with a dimension-aware selective fusion structure, termed FDPN-DASI, is designed to strengthen multi-scale feature interactions. In addition, an adaptive-threshold focal loss (ATFL) is introduced to improve the learning capability for hard samples. For efficient deployment, the LAMP pruning algorithm is further improved, and an entropy-guided entropy-adaptive magnitude-based pruning method (EA-LAMP) is proposed to enable adaptive allocation of pruning ratios across different network layers. Moreover, BCKD knowledge distillation is applied after pruning to mitigate the accuracy degradation caused by model compression. Experimental results indicate that the proposed lightweight YOLO11-SFA+EA+BCKD framework achieves a precision of 92.4%, a recall of 88.5%, and an mAP50 of 93.3%, while maintaining a compact model size of 1.6 M parameters and 4.5 G FLOPs. Compared with the baseline model, the proposed method improves precision, recall, and mAP50 by 5.9%, 5.0%, and 4.7%, respectively, while reducing the number of parameters, FLOPs, and model size by 1.0 M, 1.8 G, and 2.1 M, respectively. These results suggest that the proposed framework can improve detection performance while reducing model complexity under the current experimental setting, indicating its potential for lightweight drainage pipe defect detection tasks. Full article

Real-time scheduling in large-scale electric vehicle charging stations is challenged by stochastic vehicle arrivals, dynamic departures, limited charging resources, and station-level power constraints. To address this problem, this paper proposes a two-stage Supervised Service Matching and Reinforcement Power Dispatch (SMPD) framework, termed SMPD, which decomposes the original coupled scheduling problem into supervised service matching and reinforcement learning-based power dispatch. In the first stage, a supervised matching network learns EV-charger service suitability from historical charging-session records and determines service access decisions for feasible EV–charger pairs. In the second stage, a Soft Actor-Critic-based controller allocates continuous charging power to connected EVs under EV-side charging limits, charger capacity constraints, and the station-level total power constraint. The proposed framework is evaluated using public charging-session data from the ElaadNL dataset. Experimental results show that SMPD achieves lower average waiting time, higher average revenue, lower composite penalty, and comparable demand satisfaction compared with rule-based, single-stage reinforcement learning, and multi-agent baselines. Sensitivity and robustness analyses further indicate that SMPD maintains favorable scheduling performance and acceptable online decision time under the tested charger-scale settings and operational disturbance scenarios. These results suggest that the proposed two-stage design provides an effective and computationally tractable approach for real-time scheduling in large-scale EV charging stations. Full article

Water quality monitoring is a fundamental element of sustainable aquaculture management, as changes in parameters of physicochemical and biological properties directly affect the health, growth performance, and productivity of the aquaculture systems. Although traditional machine learning (ML) methods have demonstrated effectiveness in water quality classification, their performance often depends on large amounts of labeled data, which can be challenging and expensive to collect in real-world aquaculture environments. This study explores a few-shot learning (FSL) framework for data-efficient water quality classification under limited supervision to address this limitation. Several FSL models, including prototypical networks (ProtoNet), Siamese Networks, and Matching Networks were developed and evaluated in a comparative experimental framework against the traditional machine learning classifiers logistic regression, random forest, support vector machine and extreme gradient boosting. Low-data learning scenarios were simulated using a structured episodic evaluation approach. Experimental results demonstrate FSL techniques outperform traditional machine learning methods across all evaluated scenarios. Among the tested methods, ProtoNet achieved the highest performance, attaining an accuracy of 94.46% and an ROC-AUC score of 98.65%, indicating superior discriminative capability and robustness. Siamese Networks also demonstrated competitive performance under highly constrained data conditions. Furthermore, latent-space visualization, confusion matrix analysis, paired t-test statistical analysis, and ablation studies confirmed that episodic meta-learning enables the learning of highly discriminative latent representations with strong generalization capability under limited labeled data conditions. The findings highlight that FSL provides a robust and scalable framework for intelligent water quality classification in aquaculture systems, particularly in scenarios where labeled data are scarce, offering significant potential for sustainable aquaculture monitoring applications. Full article

Accurate road-event detection and timely alert message dissemination are essential for the safety of connected and automated vehicles. In many scenarios, alert messages must reach not only nearby vehicles but also remote stakeholders, such as traffic management centers, cloud services, and infrastructure operators. This requirement motivates the adoption of cellular-based communication technologies in addition to short-range vehicle-to-everything (V2X) communications for data dissemination. In this work, we investigate vehicle-to-network-to-everything (V2N2X) communications for the dissemination of alert messages generated after the on-board detection of hazardous road events through machine learning (ML) algorithms. Although V2N2X connectivity is well suited for extending data dissemination beyond the local vehicular environment, its capability to guarantee prompt message delivery under strict latency constraints remains an open challenge, particularly when ML inference is integrated into the end-to-end processing pipeline. To address this issue, we develop and experimentally evaluate a proof-of-concept (PoC) platform that combines real-time road-event detection with relevant message dissemination towards both nearby and remote recipients. The proposed framework leverages 5G connectivity and publish/subscribe messaging protocols. The experimental results showcase that dissemination latency is highly influenced by both the adopted type of 5G deployment (private versus commercial networks) and the load conditions at the message broker. Full article

Smart energy meters are widely deployed in modern distribution networks, extending their role beyond revenue billing to real-time monitoring and data-driven smart city applications. However, conventional legal metrology frameworks rely on periodic recalibration and are not intended for the detection of accuracy drift or unexpected malfunctions between scheduled inspections. In scientific publications, various techniques for remote smart meters’ error surveillance are presented, but experimental verification on real distribution network data remains limited. The objective of this study is to experimentally verify two previously proposed power event-driven methods for remote estimation of active power measurement error in individual consumer meters, using a feeder-level sum meter as a reference instrument. One-second resolution electrical readings were collected from a real low-voltage distribution branch using ESP32-based local adapters communicating via MQTT over Wi-Fi, with SNTP-based clock synchronization for power event correlation. Under optimized detection parameters, the linear regression method achieved 0.20% RMSE and 0.75% maximum absolute error, and the neural network method 0.09% RMSE and 0.31%, confirming suitability for Class 1 m accuracy surveillance. Feasibility analysis of three MQTT-based deployment scenarios demonstrates that binary encoding limits local adapter buffers to 2.8 kB and worst-case daily channel demand to 2000 kB, confirming the practical viability of the proposed architecture. Full article

In complex urban environments, GNSS satellite signals suffer from severe multipath errors caused by building occlusion and reflection, which significantly degrades the accuracy of precise point positioning (PPP). This paper proposes a deep-learning-based multipath hemispherical grid correction model (DL-MHM) that integrates combined filtering and satellite embedding mechanisms. The model adopts the multi-system interoperable MHM framework to achieve effective multipath error correction. First, pseudorange and carrier phase observation residuals are calculated using the ionosphere-free combination for PPP. Then, a joint median and Kalman filtering scheme is applied to suppress noise in multi-day continuous residual sequences. A transformer-based time-series learning model is constructed, which introduces satellite-specific embedding vectors to characterize the differences between individual satellites and deeply fuse temporal features. This enables the model to adaptively fit the residual variation patterns of different satellites and accurately extract multipath errors. Finally, the multipath components predicted by the deep learning model are incorporated into the multi-system interoperable MHM model to generate the final multipath corrections. Test results show that in heavily obstructed urban scenarios, the root mean square (RMS) values of the east (E), north (N), and up (U) coordinate residuals are improved by 49.27%, 1.80%, and 3.35%, respectively, after DL-MHM correction compared to the uncorrected data. In open-sky environments, the corresponding improvements are 7.70%, 5.48%, and 34.28%. In all experimental scenarios, the proposed method outperforms both the conventional multipath hemispherical map (MHM) model and the convolutional neural network-long short-term memory (CNN-LSTM)-based MHM model in terms of overall multipath correction performance. The experimental results demonstrate that the proposed DL-MHM model can effectively mitigate multipath errors in complex urban scenarios and significantly improve the accuracy of GNSS precise positioning. Full article

The increasing penetration of distributed generation (DG) in distribution systems poses significant operational challenges, including increased power losses, voltage profile deviations, and variations in short-circuit currents. These issues may compromise network safety, reliability, and the selectivity of protection schemes under different operating scenarios. This paper proposes a hybrid optimization methodology for the optimal placement and sizing of DG, aiming to minimize active power losses while ensuring voltage regulation and keeping short-circuit currents within permissible limits. An integrated approach is proposed that combines a mesh-to-radial network reconfiguration strategy with a modified Ant Lion Optimization algorithm, known as ALO-DG, enabling the simultaneous optimization of network topology and the allocation of distributed generators at candidate buses. The problem is formulated taking into account power balance constraints, voltage limits, distribution network capacity limits, and short-circuit current limits. The proposed methodology achieved substantial reductions in active power losses in the IEEE 33-bus and 69-bus test systems, reaching 84.42% and 91.56%, respectively. These improvements were accompanied by enhanced voltage profiles while preserving the radial operating structure of the distribution networks. Furthermore, the proposed hybrid methodology serves as a tool for the planning and operation of distribution systems with high DG penetration, particularly in scenarios where grid security and protection coordination are critical considerations. Full article

Face super-resolution (FSR) aims to reconstruct high-quality high-resolution face images from low-resolution inputs. Although CNN–Transformer hybrid models have shown promising performance by jointly modeling local textures and global dependencies, their large parameter sizes and high computational costs hinder practical deployment in resource-constrained scenarios such as mobile devices and embedded systems. Meanwhile, existing lightweight SR models usually reduce complexity by simplifying network depth, channel dimensions, or convolutional operations, which may weaken feature representation capability and lead to insufficient recovery of fine facial structures. To address these issues, this paper proposes HCTIUNet, a lightweight CNN–Transformer hybrid network based on an inverted U-shaped architecture. Specifically, the proposed network integrates lightweight CNN branches for local facial texture extraction and Transformer branches for global dependency modeling, while introducing a multi-scale feature interaction strategy and a global feature refinement module to enhance facial structural details. Experimental results on the FFHQ, CelebA, and Helen datasets demonstrate that HCTIUNet achieves competitive performance under the ×8 face super-resolution setting, obtaining PSNR/SSIM/LPIPS values of 27.55 dB/0.765/0.225, 27.63 dB/0.761/0.212, and 27.53 dB/0.777/0.213, respectively. Moreover, HCTIUNet contains 10.5 M parameters, requires 9.9 G FLOPs, and achieves an inference time of 0.021 s. These results indicate that the proposed method achieves a favorable trade-off between reconstruction accuracy, perceptual quality, and computational efficiency, making it suitable for efficient face super-resolution applications. Full article